Building Reliable AI Infrastructure at Scale with Extreme Co-design

Reliable AI infrastructure at scale starts long before the first system arrives. In this session, learn how NVIDIA and CoreWeave are applying extreme co-design across compute, networking, software, and operations to bring up next-generation AI factories built for the NVIDIA Vera Rubin platform. We'll explore how Vera CPUs, Rubin GPUs, NVLink, ConnectX, BlueField, and Spectrum-X Ethernet work together to deliver predictable performance, resiliency, and efficiency for model builders and AI-native companies serving production AI applications.